Add prior_lea argument to switch and if_ methods

In some situations, several entries in CPT are unknown but a prior (unconditional) probabilities are knwon. Lea provides the revised_with_cpt for such case.

c = event(0.034)
a1 = event(0.92).revised_with_cpt((c, event(0.19)))
print (P(a1))
# -> 0.92
print (P(a1.given(c))
# -> 0.19

However this method, beside being not very intutive, produces a Blea instance (clause-oriented CPT), which may be inefficient.

As alternative, it is proposed to add an optional prior_lea argument to switch method. This argument should be an alternative to the optional default_lea argument. prior_lea defines the prior unconditional probabilities, from which the default_lea can be calculated. The same optional argument can be added to if_ method, as alternative to else_lea. Then, the following defintions shall be functionally equivalent to the previous one (althouth they produce Tlea instances instead of Blea instances):

a2 = c.switch({True:event(0.19)}, prior_lea=event(0.92))
print (P(a2))
# -> 0.92
print (P(a2.given(c)))
# -> 0.19
a3 = if_(c, event(0.19), prior_lea=event(0.92))
print (P(a3))
# -> 0.92
print (P(a3.given(c)))
# -> 0.19

Comments (5)